Merge pull request #298563 from craigshoemaker/aca/cold-start-updates

[Container Apps] Update: Cold start

Author: v-shils

articles/container-apps/TOC.yml

@@ -78,8 +78,6 @@
       href: https://youtu.be/OxmVds31qL8
     - name: Communicate between multiple apps
       href: connect-apps.md
-    - name: Reduce cold-start times
-      href: cold-start.md
 - name: Develop & deploy
   items:
     - name: Deployment options
@@ -225,6 +223,8 @@
       displayName: Scale rules and performance overview
     - name: Scale a container app
       href: tutorial-scaling.md
+    - name: Reduce cold-start times
+      href: cold-start.md
 - name: Domains & certificates
   items:
   - name: Overview

articles/container-apps/cold-start.md

@@ -5,7 +5,7 @@ services: container-apps
 author: craigshoemaker
 ms.service: azure-container-apps
 ms.topic: conceptual
-ms.date: 04/15/2025
+ms.date: 05/06/2025
 ms.author: cshoe
 ---
 
@@ -27,12 +27,11 @@ Often images transition training to inference usage with only minimal tweaks. Ma
 
 Use container registries close to your Container Apps environment. Usually this means you want to use an Azure Container Registry deployed in the same region as your environment, or a premium registry that features global distribution.
 
-## Use artifact streaming
+## Manage large downloads
 
-Azure Container Registry [offers image streaming](/azure/container-apps/serverless-gpu-nim#enable-artifact-streaming-recommended-but-optional) which can speed up image startup by up to 15%.
+Use [storage mounts to hold critical data close to your container app](storage-mounts.md), especially when the file sizes are large. For instance, if your app requires a large language model, you can pre-download the model to your storage account. By reading large files from a storage account, you avoid the latency of pulling files across the internet.
 
-> [!NOTE]
-> Artifact streaming is only supported with [serverless GPUs](./gpu-serverless-overview.md).
+If you are creating a storage mount for AI workloads, make sure you use the most appropriate [mount options](/troubleshoot/azure/azure-kubernetes/storage/mountoptions-settings-azure-files) for your needs.
 
 ## Implement custom liveness health probe or start listening early
 

articles/container-apps/gpu-serverless-overview.md

@@ -7,7 +7,7 @@ ms.service: azure-container-apps
 ms.custom:
   - ignite-2024
 ms.topic: how-to
-ms.date: 03/17/2025
+ms.date: 04/24/2025
 ms.author: cshoe
 ---
 
@@ -93,24 +93,11 @@ Serverless GPUs are run on consumption GPU workload profiles. You manage a consu
 
 ## Improve GPU cold start
 
-You can improve cold start on your GPU-enabled containers by enabling artifact streaming on your Azure Container Registry. For more information, see [enable artifact streaming](/azure/container-registry/container-registry-artifact-streaming?pivots=development-environment-azure-cli).
+You can significantly improve cold start times by enabling artifact streaming and locating large files, like large language models, in a storage mount.
 
-> [!NOTE]
-> To use artifact streaming, your container images must be hosted in a premium Azure Container Registry.
-
-Use the following steps to enable image streaming:
-
-1. Open your Azure Container Registry in the Azure portal.
-
-1. Search for **Repositories**, and select **Repositories**.
-
-1. Select your repository name.
-
-1. From the *Repository* window, select **Start artifact streaming**.
-
-1. Select the image tag that you want to stream.
+- [Artifact streaming](/azure/container-registry/container-registry-artifact-streaming): Azure Container Registry offers image streaming which can significantly speed up image startup times. To use artifact streaming, your container images must be hosted in a premium Azure Container Registry.
 
-1. In the window that pops up, select **Create streaming artifact**.
+- [Storage mounts](cold-start.md#manage-large-downloads): Reduce the effects of network latency by storing large files in an Azure storage account associated with your container app.
 
 ## Submit feedback